Institute of research into the fundamental laws of the Universe

The STEREO experiment aims to test the existence of a sterile neutrino. For physicists, this new type of neutrino could be the best chance to catch a glimps of the beyond (that is beyond the standard model). It is not easy, however, to hunt down this exotic particle that frees itself from all interactions with matter. The key to its detection lies in its expected mix with other neutrinos, a well-established property for the three active neutrinos already known, whose discovery earned the Nobel Prize for physics in 2015. For that reason, STEREO is located closest to a very intense source of neutrinos: 10 m from the research reactor of the ILL, Grenoble. At this distance, about 4×10¹⁵ neutrinos, resulting from the beta decays of the fission products, pass each second through the 2 m³ of scintillator liquid of the detector (Figure 1) for only one neutrino intercepted every ... 4 minutes. If the sterile neutrino exists, it will induce an oscillation in the number of active neutrinos detected, visible by comparing the spectra measured in the 6 cells of STEREO.

With such a low rate of neutrinos, one must be able to reject all other signals (natural radiation, reactor activity) that could mimic the signal of a neutrino. Fortunately, physicists have a rather unique signature of the interaction of a neutrino with a proton of the detector: the detection of a positronⁱ followed by the capture of a neutron (Figure 2). To make this process more efficient, gadolinium (Gd) is blended with the scintillator liquid. This element holds the record of appetite for neutrons. In just a few microseconds, it will capture the neutron produced by a neutrino and emit a cascade of gamma rays to signal its capture with a total energy of 8 MeV, well above most disturbing signals.

The oscillations sought by STEREO develop at the scale of meters. The detector is therefore compact and a non-negligible part of the gammas of the cascade will escape the liquid scintillator. The nice signal expected at 8 MeV will thus be adorned with a broad low energy tail, filled with all these partial energy deposits of the cascade. A detection cutoff is applied to stay above the background (typically, energies above 4.5 MeV are selected). But STEREO wants to control its detection efficiency to the %-level, so it is necessary to describe very precisely the gamma cascades of Gd.

At this point FIFRELIN, being well acquainted with the complex terrains of gamma cascades, enters the stage. It is a Monte Carlo code developed at CEA / DEN Cadarache that simulates the production and de-excitation of fission fragments to meet nuclear data needs for reactor physics [2]. In particular, FIFRELIN is capable of modelling the emission cascade of gammas and electrons resulting from the de-excitation of a nucleus created by neutron capture. To do this, it uses all available nuclear structure data that describes the first excited levels. But after absorption of a neutron, the excitation energy of the Gd nucleus reaches a continuum of levels (Figure 3). FIFRELIN is then based on level-density models to complete the level schemes. After calculating all the probabilities of partial inter-level transitions, the code samples millions of electromagnetic cascades, while simultaneous controlling the number and energy of the gamma rays. These cascades are then used in the simulation of the STEREO detector response.

Figure 4 illustrates the improvement obtained, thanks to FIFRELIN, in the description of the energy measured after the neutron captures. Capture peaks on Gd (8 MeV) and to a lesser extent on hydrogen (2.2 MeV) are clearly visible. The initial agreement obtained with the GEANT4 simulations (top graph) seems satisfactory, but the residual distortions were nevertheless sufficient to generate 4.5% difference between the simulated and measured detection efficiencies for a neutron source in the center of a cell. Thanks to the FIFRELIN simulations (bottom graph), the agreement becomes almost perfect. This is the case for the alignment of the peaks as well as the distribution of intermediate energies, very sensitive to the description of the cascades. The agreement between simulation and data reaches down to only 0.5%, with an associated sub-% level uncertainty [1].